KR101208171B1 - Interference presubtracting method and trasmitter using the same - Google Patents

Abstract

The transmitter includes a channel encoder performing channel coding on an information symbol, an interleaver interleaving the channel coded signal, an interference deduction unit for subtracting known interference from the interleaved signal, and the shape of the signal set from which the interference has been removed. A source encoder for generating a transmission signal. DPC can be implemented by adding only the Viterbi algorithm to the transmitter. The receiver can use a turbo decoder, which is generally used for decoding, almost intact.

Interference Prepayment, DPC, Turbo Code, Multi-User, MIMO

Description

Interference presubtracting method and trasmitter using the same

1 is a block diagram illustrating a transmitter and a receiver according to an embodiment of the present invention.

2 is an exemplary view showing an example of a lattice using a repetition code.

3 is a block diagram modeling the transmitter of FIG. 1.

5 is a block diagram of an equivalent model of the transmitter of FIG. 4.

6 is a block diagram of an equivalent model of the transmitter of FIG. 5.

7 is a block diagram illustrating a transmitter to which an equivalent model is applied.

8 is a block diagram illustrating a decoding portion of a receiver.

9 is a block diagram illustrating the decoding portion of FIG. 8 in terms of iterative decoding.

10 is an exemplary diagram showing a graph of source code.

11 is an exemplary diagram illustrating a notation of an EXIT chart.

12 is a graph illustrating an EXIT chart according to a simulation.

13 is a graph showing a simulation result for the interleaver effect.

14 is a graph showing the capacity of the DPC system according to the present invention and the prior art and the DPC system according to the present invention.

** Explanation of symbols in main part of drawing **

110: channel encoder

120: interleaver

150: interference deduction

160: source encoder

The present invention relates to wireless communication, and more particularly, to an interference preduction method and a transmitter using the same.

The demand for communication services is rapidly increasing, including the generalization of information and communication services, the emergence of various multimedia services, and the emergence of high quality services. Various wireless communication technologies have been studied in various fields to satisfy these demands.

Recently, research based on point-to-point communication by single users has been expanded to point-to-mulitpoint communication by multiple users. In the one-to-many communication, an interference presubtraction method in a transmission end is drawing attention. Signal A is called a signal to send to user A, and signal B is called a signal to send to user B. Signal A is first processed from the proper association with signal B to produce a noise-like signal (A '), which is added to signal B and sent to the channel. Receiving this signal, user B considers and decodes both the noise from the channel and the processed signal A 'in the original signal B as noise. User A can completely recover signal A from the processed noise A '. It is sometimes called dirty paper coding (DPC) because any processing on the original signal is like writing on dirty paper and the receiving end can restore the original text no matter how dirty the paper is. Examples of DPCs are described in M. Costa, "Writing on Dirty Paper", IEEE Trans. on Inf. Theory, vol. IT-239, No. 3, May 1983. In this document, Costa shows that under DPC, the channel capacity is the same as in the absence of interference.

The DPC may be referred to as an interference signal canceling technique in a transmitting end that prevents the receiving end from being affected by the interference signal when the transmitting end knows the interference signal in advance in the presence of an interference signal in addition to the noise signal in the channel.

Two error correction codes may be used as the DPC implementation structure. One is an error correction code, and the other is a shaping code that changes the shape of the transmission signal set. To approach the ideal channel capacity, a good error correction code and a high gain formal code should be used. In order to guarantee this performance, a long codeword and a complicated structure of a transceiver are required. In addition, such a DPC system is difficult to implement in a real environment because it takes a lot of time in the encoding and / or decoding process.

DPCs can be thought of as error correction codes with the ability to eliminate interfering signals. Although the DPC guarantees high data rates in a multi-user multi-input multiple output (MIMO) environment, it is necessary to change the part of the error correction code in order to actually reflect the DPC in the system, such as a turbo code. It is difficult to guarantee backward compatibility with systems using existing error correction codes. It is easy to upgrade and advantageous in terms of system recycling that can be implemented with only minor modifications in existing error correcting systems.

The technical problem to be achieved by the present invention is to provide an interference pre-deduction method and a transmitter using the same to reduce the complexity of the DPC implementation.

Another object of the present invention is to provide an interference preduction method for increasing backward compatibility and a transmitter using the same.

In accordance with an aspect of the present invention, a transmitter includes a channel encoder for performing channel coding on an information symbol, an interleaver for interleaving the channel coded signal, an interference deduction unit for subtracting known interference from the interleaved signal, and the interference cancellation. And a source encoder for changing a shape of the signal set to generate a transmission signal.

An interference preemptive method according to another aspect of the present invention convolutionally encodes information bits to generate an error correction code, and interleaves the error correction code. Pre-interference is deducted from the interleaved signal. Then, the interference-deducted signal is convolutionally encoded to change the aggregate shape of the transmission signal.

According to another aspect of the present invention, a source decoder receiving a received signal and a virtual input to a receiver for receiving a transmission signal through which a known interference is removed through a channel. And a channel decoder for performing channel decoding and an interleaver disposed between the source decoder and the channel decoder, wherein the received signal is repeatedly decoded through the source decoder, the interleaver, and the channel decoder.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Like reference numerals designate like elements throughout the specification.

1 is a block diagram illustrating a transmitter and a receiver according to an embodiment of the present invention.

Referring to FIG. 1, the transmitter 100 includes a channel encoder 110, an interleaver 120, an interference presubtracter 150, and a source encoder 160.

The channel encoder 110 encodes the input information bit through the channel code. The channel code functions to correct an error due to the channel, and the channel encoder 110 may be a convolutional encoder. The convolutional encoder convolutionally encodes the information bits to produce a codeword, which is an error correction code.

The interleaver 120 interleaves the codewords output through the channel encoder 120. Since the convolutional code is used as the channel encoder 110 and the shaping part after the interleaver 120 has the form of the convolutional code, the interleaver 120 is disposed between the turbo 100 at the transmitter 100. It was considered to produce a form similar to the trubo code.

The up-sampler 130 creates a repetition code in which the interleaved signal is repeated. The repetition code is a repetition of the input signal by two or more times the length of the input signal. The mapper 140 constellation maps the repetition code. There is no limitation on the mapping method, and m-quadrature phase shift keying (m-PSK) or m-quadrature amplitude modulation (m-QAM) may be used. In the following description, QPSK is used for clarity.

The reason why the repetitive code is generated through the upsampler 130 is that if the repetitive code is not used, the point on the lattice Λ and the QPSK generated by the source coding described later, for example, a modulo operation, will be described. This is because constellations mapped by each other overlap each other and the receiver 200 cannot accurately recover information. Mapping with repetitive codes can be said to act like an inverse syndrome former.

2 is an exemplary view showing an example of a lattice using a repetition code.

Referring to FIG. 2, a modular operation is required to implement a DPC. When the message sent to the user is called w and the interference previously known by the transmitter 100 is S, the easiest way to implement the DPC is to make the transmission signal X X = wS. In this case, the received signal Y at the receiver 200 becomes Y = X + S + N = (wS) + S + N = w + N (where N is noise). However, this method has some problems. First, if the interference S is large, the power E [X ² ] = E [w ² ] + E [S ² ] of the transmission signal X exceeds the specified transmit power constraint. Secondly, when the DPC is used in a multi-user MIMO system, since the interference S is a signal of another user, other users may not receive any signal when the transmitter 100 directly performs a process such as X = WS. Therefore, this method is not suitable.

For any interference S, it is necessary to change the shape of the transmitted signal set in order for the DPC to operate. Modulo operations are performed for this purpose. For example, when a QPSK symbol has a form of ± 1 ± j in a constellation using QPSK, modulo operation allows -1 + j to be -5-3j, -5 + j, + 3 + j, +3+. 5j and the like operate as if they are the same symbol to be prepared for any interference S.

However, if the transmission is performed through a simple modulo operation, if the transmission signal is an n-dimensional codeword, it has an n-dimensional cube shape, which is inefficient in terms of power. This is because transmitting a point near an edge consumes more power than other points.

Based on a grid of all points mapped through modulo operations, a set of regions, or regions, that map to one lattice point when all points in two-dimensional space are mapped to the nearest lattice point. The shape of becomes a two-dimensional rectangle (usually an n-dimensional cube). A grid is a set of discrete points distributed in real space. The area underlying this lattice is called the fundamental Voronoi region.

To make this concept more general, all vectors x in R ⁿ can be selected as shown in Equation 1, provided that the lattice Λ satisfies Λ⊂ R ⁿ (n-dimensional real vector).

The set of points closer to zero (the origin) than any other λ in R ⁿ is called the basic Voronoi region υ. This can be expressed as r = x mod Λ. Where r = x mod Λ = x-Q (x) Q (x) is the n-dimensional point closest to x among λ∈Λ. Since the shape of the basic Voronoi region determines the power of the transmission signal, if the modulo operation is performed based on a well-designed lattice Λ, the shape of the basic Voronoi region is closer to the sphere shape. Can overcome.

Multiplying the message w = {0,1} by the generating matrix G = [1 1] may have two codewords {[0 0], [1 1]}. A lattice Λ made using this is shown in FIG. 2. In this case, the basic Voronoi region may be displayed as the region shown in FIG. 2.

Referring to the repetition code, [1 1] and [0 0], [1 0] and [0 1] are treated as the same points in the information on which the grid Λ-based mod Λ operation shown in FIG. 2 is performed. That is, [0 0] mod Λ = [1 1] mod Λ = [0 0] and [1 0] mod Λ = [0 1] mod Λ = [1 0]. This means that only half of the dots can be used to map the constellation. Therefore, iterative code solves this problem.

Referring back to FIG. 1, the interference deduction unit 150 deducts a dithering signal U and a value obtained by multiplying the interference S known by the transmitter 100 by the scaling facor α with respect to the mapped symbol w. . In other words, the output q of the interference subtraction unit 150 is q = w-αS-U.

The dither signal U is a uniform random variable in the basic Voronoi region v known to both the transmitter 100 and the receiver 200. When the interference S is biased in one place without being sufficiently random, the transmission signal is not uniform in the basic Voronoi region v. In this case, when the user decodes his signal, the interference, which he considers to be noise, is skewed to one place, so that the original information cannot be properly restored. Therefore, the dither signal U makes the transmission signal X uniform in the basic Voronoi region υ. In addition, the dither signal U makes the message w and the transmission signal X independent of each other when passing through the channel. For example, the dither signal U is a uniform real value at (-1,1] ^N.

The magnification α is a value known to both the transmitter 100 and the receiver 200. This is set to minimize the effective noise described later. For example, α = SNR / (1 + SNR) can be set. Where SNR is the signal-to-noise ratio obtained through the channel.

The source encoder 160 modulates the output of the interference deduction unit 150 and outputs the transmission signal X. FIG. The source encoder 160 performs a modulo operation to change the shape of the set of transmission signals x. That is, the transmission signal X output from the source encoder 160 is represented by Equation 2 below.

mod Λ selects the closest point among the grid points created by lattice Λ and outputs the difference from the input value to the output. As will be described later, the proposed system uses both the channel code and the source code as convolutional codes, and there is an interleaver between the two, so that the decoding process is similar to the existing turbo code.

The source encoder 160 includes a processor 161 that performs a Viterbi algorithm to efficiently implement modulo operations. The source encoder 160 is supplied with an input of a signal q, which is a mixture of a QPSK mapped signal w, an interference S, and a dither signal U. The purpose of mod Λ is to find the codeword c closest to the input q of all points of the grid Λ. However, since the number of all points of the grid Λ is infinite, the following work is required.

를 바탕으로 한다. 처리기(161)는 비터비 알고리즘을 통해 점 c_i를 찾고, 소스 인코더(160)는 비터비 알고리즘으로부터 찾은 점 c_i를 입력 신호 q로부터 빼서 전송신호 X를 만든다.In order to satisfy the above equation, the assumption of a lattice chain Z ⁿ / Λ / 2Z ⁿ is required. Therefore, we process q mod 2Z ⁿ for input q and mod Λ, but find the codeword closest to q mod 2Z ⁿ among the finite codewords formed by lattice points or convolutional codes in the basic Voronoi region. Find and send the difference to it and the shaping process ends. The Viterbi algorithm is an algorithm that efficiently finds the most likelihood codewords produced by convolutional codes. When calculating the internal metic

Based on The processor 161 finds the point c _i through the Viterbi algorithm, and the source encoder 160 subtracts the point c _i found from the Viterbi algorithm from the input signal q to produce the transmission signal X.

Meanwhile, the receiver 200 includes a source decoder 250, an interleaver 260, and a channel decoder 270. The transmission signal X passes through the channel, where noise N and interference S are added to the input signal R = X + S + N of the receiver 200. The scaling unit 210 multiplies the input signal R by the magnification α. The dither adder 220 adds the dither signal U to the input signal multiplied by the magnification α. The modulo operator 230 performs a modulo operation on the output of the dither adder 220 and outputs a received signal Y as shown in Equation 4 below.

The effective noise N 'can be obtained from the above equation. That is, since R = X + S + N, Y = (w-w + αX + αS + αN + U) mod Λ = (w + αX + (αS + Uw) + αN) mod Λ = (w + αX + (αS + Uw) mod Λ + αN) mod Λ = (w + αX + (-X) + αN) mod Λ = (w + (1-α) X + αN) mod Λ = (w + N ') mod Λ. At this time, the transmission signal X becomes independent of w due to the dither signal U, which is equal to the effective noise Z '= (1-α) U + αN. The magnification α can be set to minimize the effective noise N '.

The received signal Y is demapped by the demapper 240, input to the source decoder 250, and restored to the original information bits according to an iterative decoding method between the interleaver 260 and the channel decoder 260. do.

The decoding method will be described below. To describe the decoding method, an equivalent model of the encoding method is used.

3 is a block diagram modeling the transmitter of FIG. 1.

Referring to FIG. 3, the signal q minus the known interference signal S and the dither signal U is applied to the input of the source encoder 160 to the signal w passed through the channel encoder 110 and the interleaver 120. Here, the processor 161 finds and outputs the point closest to the input q among all the points of the lattice Λ.

4 is a block diagram of an equivalent model of the transmitter of FIG. 3.

Referring to FIG. 4, the transmitter 300 uses a modulo operator 355 to modulate a signal q obtained by subtracting the interference signal S and the dither signal U, which are previously known to the signal w through the channel encoder 310 and the interleaver 320. Perform the operation with Even if a modulo operation is performed on the signal q at the front end of the processor 361, an equivalent relationship is established with the transmitter of FIG. 3 which does not perform a modulo operation. This is because finding the closest point to all points of the grid Λ spread everywhere in the n-dimensional plane is the same as finding the closest point only in the region after mod 2Z. This is because q mod Λ = (q mod 2Z) mod Λ is satisfied because the lattice Λ satisfies the lattice relation Z / Λ / 2Z. For example, (5.9) mod 2Z = -0.1 and (5.9 mod 4Z) mod 2Z = (1.9) mod 2Z = -0.1 for a grid satisfying the lattice relation Z / 2Z / 4Z.

5 is a block diagram of an equivalent model of the transmitter of FIG. 4.

Referring to FIG. 5, the transmitter 400 is an equivalent model of the transmitter 300 of FIG. 4. In addition, since the transmitter 300 of FIG. 4 is an equivalent model of the transmitter 100 of FIG. 3, the transmitter 400 is an equivalent model of the transmitter 100 of FIG. 3. The power of the signal X transmitted to the channel by the processor 361 of the transmitter 300 of FIG. 4 is minimized. This is because the signal c _i closest to the signal q is found, and the difference between q and c _i becomes the transmission signal X. Thus, there is some virtual input equivalent to sending the codeword obtained through the source encoder 460 minus q mod 2Z. The subtraction operation is also the same as the addition operation under mod 2Z.

Accordingly, modulo operation is performed on the signal q by the first modulo operator 455, and the adder 457 adds a value obtained by source encoding the virtual input to the output value of the first modulo operator 455. Subsequently, if the output of the adder 457 is modulated by the second modulo operator 470, the transmitter 400 becomes an equivalent model of the transmitter 300 of FIG. 4. Here, the source encoder 460 may be a convolutional encoder.

6 is a block diagram of an equivalent model of the transmitter of FIG. 5.

Referring to FIG. 6, the transmitter 500 is an equivalent model of the transmitter 400 of FIG. 5. The sum of the two binary bits under the mod 2Z operation is the same as the exclusive OR (XOR) operation. For example, (1 + 1) mod 2 = 0 = XOR (1,1) and (1 + 0) mod 2 = 1 = XOR (1,0). Accordingly, in the transmitter 400 of FIG. 5, the first modulo operator 455 and the adder 457 are equivalent to the XOR operator 530.

Since the output value of the source encoder 560 does not interfere with where the interference signal? S and the dither signal U are subtracted, the transmitter 500 is an equivalent model of the transmitter 400 of FIG.

7 is a block diagram illustrating a transmitter to which an equivalent model is applied.

Referring to FIG. 7, the information bits are repeated coded by the upsampler 630 via the channel encoder 610 and the interleaver 620. The virtual input is encoded by the source encoder 660 to perform an XOR operation with each iteration code by the XOR operator 632 and the second XOR operator 634. The XOR-calculated signal is constellation mapped by the mapper 640, and the interference deduction unit 650 deducts the interference signal αS and the dither signal U. The interference-deducted signal is mod 2Z performed by the modulo operator 662 to become the transmission signal X. The channel encoder 610 and the source encoder 660 may both be convolutional encoders.

8 is a block diagram illustrating a decoding portion of a receiver.

Referring to FIG. 8, the source decoder 750 and the channel decoder 770 respectively correspond to the source encoder 660 and the channel encoder 610 in the transmitter 600 of FIG. 7. The source decoder 750 and the channel decoder 770 use a soft input soft output decoder using the BCJR algorithm. In general turbo code, two BCJR decoders repeatedly exchange information based on soft information received from a channel to increase reliability of information bits. In the present invention, since the convolutional code is used for both the channel code and the source code, the BCJR decoder can be used as the decoder. The BCJR algorithm is described in L.R. Bahl, J. Cocke, F. Jekinek, J. Raviv, "Optimal Decoding of Linear Codes for Minimizing Symbol", IEEE Trans. Info. Theory, Vol IT-20, pp. 284-287, March 1974.

The difference between the decoding part and the general turbo decoder is that the received signal Y is a modulo operation on the signal coming through the channel, and thus it is truncated even though the noise form is AWGN (additive white Gaussian noise). In addition, the log likelihood ratio (LLR) value for the virtual input to the source encoder 660 is always set to zero. This is because the receiving end does not know which virtual input is used for source coding at the transmitting end. The receiver sets LLR to 0 on the basis that all virtual inputs are available at the transmitter. The rest of the iteration and BCJR algorithm are identical to the normal turbo decoding process.

9 is a block diagram illustrating the decoding portion of FIG. 8 in terms of iterative decoding.

Referring to FIG. 9, the source decoder 850 receives a received signal Y and a virtual input, and recovers information bits by iterative decoding between the interleaver 860 and the channel decoder 870. The channel decoder 870 operates in the same way as general channel decoding. The BCJR algorithm calculates the LLR for the information bits based on the extrinsic information calculated from the source decoder 850.

The source code recognizes the existing convolutional code and the repetition after the interleaver 860 as a trellis structure at once. Otherwise, no iteration starts, no matter how good the channel is.

10 is an exemplary diagram showing a graph of source code.

Referring to FIG. 10, this considers the source code and the repetition part as one trellis structure. Information bits and virtual bits are input to the trellis structure, and what is observed on the channel is the output of the trellis structure. The virtual bit at the transmitting end is determined for the purpose of minimizing the transmission power, but the receiving end may not get any information about it. Therefore, as described above, when the BCJR algorithm is executed in the source code, the LLR value for the virtual bit is set to zero. That is, Pr (I _{virtual bit} = 0) = Pr (I _{virtual bit} = 1) = 0.5.

The updated external information is sent back to the channel decoder 870 through the BCJR algorithm by using the observed value of the channel and the external information calculated by the channel decoder 870 during the repetition process. This external information is communicated to the channel decoder 870 via the interleaver 860. An interleaver 860 is disposed between the source decoder 850 and the channel decoder 870 to eliminate the cycle of the graph of the entire system as much as possible. The interleaver 860 may use any interleaver, or may use an interleaver using a row-column exchange method.

According to the present invention, a decoding method similar to the turbo code is used. This is to increase the reliability of information through repetition. The following analysis of this iteration and the performance of the overall system is shown.

11 is an exemplary diagram illustrating a notation of an EXIT chart.

Referring to FIG. 11, one of a method of checking whether an error probability converges to 0 in the repetition method is EXIT (extrinsic information transfer) chart analysis. I _E1 is mutual information about the updated LLR calculated by the channel decoder 13, I _E2 is mutual information about the updated LLR calculated by the source decoder 11, and I _A1 is the channel. Reciprocal information about the input LLR entering the decoder 13, I _A2 is the mutual information about the input LLR entering the source decoder 11.

12 is a graph illustrating an EXIT chart according to a simulation. The x-axis is IE1, IA2 (reciprocal information of the message delivered from the channel decoder to the source decoder), and the y-axis is IE2, IA1 (reciprocal information of the message delivered from the source decoder to the channel decoder). The graph without Eb / No is a graph of mutual information (I _A1 vs I _E1 ) corresponding to the channel encoder by changing the xy axis, and the rest of the graph shows the relationship between I _A2 and I _E2 under the observed I _channel according to the channel state. Indicates.

Referring to FIG. 12, a transfer function corresponding to a source code is represented according to the noise mixed in a channel, and a condition for converging to Eb / No = 2.8 dB is created. That is, at Eb / No = 2.8dB, the mutual information value according to the arrow converges to one. If the graph of the source code transfer function for Eb / No lower than Eb / No = 2.8 dB overlaps with the graph of the channel code transfer function, the mutual information value cannot converge to one. For Eb / No greater than Eb / No = 2.8 dB, the mutual information value converges to 1.

13 is a graph showing a simulation result for the interleaver effect. An interleaver is inserted between the channel decoder and the source decoder to simulate whether or not more efficient iterative decoding is achieved. The codeword length (N) is simulated by changing to 24, 804, 10004. The parameters used in the simulation are shown in Table 1 below.

Channel AWGN + interference known at Tx Interference 5 times Gaussian than noise Spectral efficiency 0.5 Bits / s / Hz Block size (codeword length, N) 24/804/10004 Interleaver design Random interleaver Source coder convolution codes (5,7 (8)) Channel coder convolution codes (5,7 (8)) Performance measure BER vs. Eb / No

Referring to FIG. 13, a bit error rate (BER) of about 2.5 × 10 ⁻⁶ is shown at Eb / No = 2.8 dB. Without the interleaver, performance tends to deteriorate. In particular, when the interleaver is not used, the longer the codeword is, the more cycles are generated in the decoding graph, which results in a structure that makes the incorrect information more reliable during the repetition process, thereby worsening the performance.

On the other hand, the convolutional code used as the source code is fixed to (5 (8), 7 (8)), the constraint length of the convolutional code corresponding to the channel code is fixed to 5, and all other trellis possible. Simulation of the structure shows that (35 (8), 12 (8)) can be reduced to Eb / No = 2.8dB, which makes the whole system work.

14 is a graph showing the capacity of the DPC system according to the present invention and the prior art and the DPC system according to the present invention.

Referring to FIG. 14, 'No Shaping' uses mod 4Z and low density parity check (LDPC) code R = 1/2 without fixing any spectral efficiency to 1 and performs QPSK modulation. It is the case of using, and shows the performance of Eb / No = 3.7dB. When the frequency efficiency is fixed at 1, the capacity limit is 0 dB.

If the frequency efficiency is fixed at 0.5, the Shannon limit is Eb / No = -0.8175dB. According to 'Brink', there is a 1.3dB difference from Sanon's limit, indicating Eb / No = 0.48dB. 'Brink' is described in U. Erez, S. Brink, "A close-to-capacity dirty paper coding scheme", IEEE Trans. Info. Theory, Vol. 51, No. 10, pp 3417-3432, Oct. See 2005. However, to achieve this performance, we have to format a very strong channel code, for example a convolutional code with an Irregular repeat accumulate (IRA) code R = 1/6, N = 360,000 and 16 states.

In the proposed system, when the convolutional code of 16 states is used as the channel code and the source code is used as the convolutional code of 4 states, the operating point shows Eb / No = 2.65dB.

The present invention may be implemented in hardware, software, or a combination thereof. (DSP), a programmable logic device (PLD), a field programmable gate array (FPGA), a processor, a controller, a microprocessor, and the like, which are designed to perform the above- , Other electronic units, or a combination thereof. In the software implementation, the module may be implemented as a module that performs the above-described function. The software may be stored in a memory unit and executed by a processor. The memory unit or processor may employ various means well known to those skilled in the art.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those skilled in the art that various changes and modifications may be made therein without departing from the spirit and scope of the invention. You will understand. Therefore, the present invention is not limited to the above-described embodiment, and the present invention will include all embodiments within the scope of the following claims.

As described above, according to the present invention, the DPC may be implemented by adding only the Viterbi algorithm to the transmitter. The receiver can use a turbo decoder, which is generally used for decoding, almost intact. Therefore, the DPC can be implemented relatively simply and is easy to apply to an actual system.

Claims (10)

A channel encoder for performing channel coding on the information symbols; An interleaver for interleaving the channel coded signal; An interference deduction unit for subtracting known interference from the interleaved signal; And a source encoder for generating a transmission signal by changing the shape of the signal set from which interference is removed. The method of claim 1, The channel encoder and the source encoder are convolutional encoders. The method of claim 1, The interference subtraction unit further subtracts a dither signal from the interleaved signal. The method of claim 1, Further comprising a mapper for constellation mapping the interleaved signal, And the interference deduction unit deducts interference from the mapped signal. 5. The method of claim 4, And the interleaved signal is repeatedly input to the mapper. The method of claim 1, The source encoder further comprises a processor that performs a Viterbi algorithm on the signal from which interference has been removed; And the transmission signal is a difference between the signal from which interference is removed and the output of the processor. Convolutionally encoding the information bits to generate an error correction code; Interleaving the error correction code; Pre-submitting known interference from the interleaved signal; And Weaving-encoding the signal from which the interference is deducted to change the aggregate shape of the transmission signal. The method of claim 7, wherein And changing the aggregate shape comprises modulating the interleaved signal as a module. In a receiver for receiving a transmission signal through which a known interference has been removed, A source decoder configured to receive a received signal and a virtual input; A channel decoder for performing channel decoding; And An interleaver disposed between the source decoder and the channel decoder, And a receiver for repeatedly decoding the received signal through the source decoder, the interleaver, and the channel decoder. The method of claim 9, And the virtual input has a log likelihood ratio (LLR) value of zero.

Images